서브메뉴
검색
Towards Risk-Aware Planning in Uncertain Environments
Towards Risk-Aware Planning in Uncertain Environments
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105223
- ISBN
- 9798291566381
- DDC
- 621
- 서명/저자
- Towards Risk-Aware Planning in Uncertain Environments
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 135 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: B.
- 주기사항
- Advisor: Vasudevan, Ram.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약Robotic manipulators have the potential to enhance modern healthcare by performing diagnostic procedures in point-of-care settings, assisting surgeons in operating rooms, and accelerating the discovery of novel therapeutics in research laboratories. However, researchers must address several key challenges to ensure robots can accomplish these important tasks. First, robots should be autonomous. This means robots should be capable of sensing their surroundings, gathering and learning from new information, and making their own decisions about how to accomplish specific tasks. Second, robots should be safe and only perform actions that are guaranteed to not damage objects in the environment, nearby humans, or even the robot itself. Third, robots should be robust to uncertainty. For example, a research robot should be able to clear instruments from a workbench without knowing the exact mass or friction coefficient of each instrument. Finally, robots should operate in real-time. This ensures that robots can quickly adapt their behavior to task or environment changes.The goal of this thesis is to address the challenges discussed above by developing a novel motion planning framework that integrates perception, reachability analysis, and control algorithms to generate safe trajectories in a receding horizon fashion. The proposed framework is the result of the following major contributions: (1) the development of a core trajectory planning framework based on reachability analysis; (2) the development a novel sphere-based safety representation that facilitates the integration of perception models into the planning framework; and, (3) the application of the trajectory planner as a novel differentiable neural network layer. My approach is distinctive in three ways. First, it ensures that both safety and dynamics constraints are robust to uncertainty and satisfied in continuous-time rather than discrete-time. Second, it studies how to integrate artificial intelligence, trajectory optimization, and control into a coherent framework for robot learning with theoretical guarantees. Third, by leveraging accelerated computer hardware, the proposed framework is computationally tractable and capable of being implemented in real-time on real robotic systems. In this thesis, we demonstrate this framework's effectiveness by solving a variety of challenging motion planning and manipulation tasks in simulation and on real hardware.The first major contribution of this thesis is the development of an algorithm called Autonomous Robust Manipulation via Optimization with Uncertainty-aware Reachability, or ARMOUR. As we show throughout this thesis, ARMOUR is the core algorithm of the proposed planning framework. The key insight behind ARMOUR is to combine classical robotics algorithms with interval arithmetic to compute reachable sets that overapproximate the behavior of the robot in continuous-time. This allows one to use the forward kinematics to overapproximate the swept volume of the robot and inverse dynamics to overapproximate its dynamics. ARMOUR integrates a technique called reachability analysis with a novel nonlinear robust controller to generate safe trajectories that a robot follows in successive steps to reach a goal state. At every planning iteration, a continuum of parameterized trajectories is used to construct reachable sets that overapproximate all of the robot's positions, velocities, and torques that can be reached from a given set of initial conditions over a specified interval of time. To account for uncertainty in the robot dynamics, the position reachable set is buffered by the robust controller's worst-case tracking error. Using the reachable sets as constraints, ARMOUR solves a nonlinear optimization problem to find a trajectory that brings the robot close to a desired goal, is guaranteed to avoid collisions with obstacles, and satisfies the position, velocity, and torque limits of the robot. ARMOUR is demonstrated to outperform other state-of-the-art motion planning algorithms such as ARMTD and CHOMP. The second major contribution of this thesis addresses two key limitations of ARMOUR by developing of an algorithm called Safe Planning for Articulated Robots using Reachability-based Obstacle avoidance With Spheres, or SPARROWS. First, the obstacle-avoidance reachable sets constructed by ARMOUR are overly conservative. This can make it difficult for ARMOUR to generate motion plans in cluttered scenes. Second, ARMOUR assumes all obstacles in the scene are known and can be modelled as convex polytopes. To overcome these limitations, SPARROWS develops a novel sphere-based representation that is shown to be far less conservative than ARMOUR. This results in more flexible, less conservative planning in cluttered environments, significantly outperforming additional state-of-the-art motion planners such as TrajOpt, MPOT, and cuRoboFinally, this thesis introduces SPLANNING, a perception-based, risk-aware trajectory planning framework. Unlike ARMOUR and SPARROWS, SPLANNING represents complex scenes as collections of normalized 3D Gaussian splats. This representation provides a smooth and fully differentiable description of the workspace, avoiding the limitations of point clouds and occupancy grids. SPLANNING combines the 3D Gaussians with the spherical reachable sets from SPARROWS to compute an upper bound on the probability of collision between the robot and its environment. This bound is then used as a constraint in a gradient-based optimization framework, enabling real-time planning in visually complex scenarios. Together, ARMOUR, SPARROWS, and SPLANNING form the first reachability-based framework that unifies perception, planning, and control into a general approach for safe and efficient manipulation under uncertainty.
- 일반주제명
- Mechanical engineering
- 일반주제명
- Robotics
- 일반주제명
- Computer engineering
- 키워드
- Robots
- 키워드
- Motion planning
- 키워드
- Convex polytopes
- 키워드
- Manipulation
- 기타저자
- University of Michigan Robotics
- 기본자료저록
- Dissertations Abstracts International. 87-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359843
■00520260202105223
■006m o d
■007cr#unu||||||||
■020 ▼a9798291566381
■035 ▼a(MiAaPQ)AAI32271825
■035 ▼a(MiAaPQ)umichrackham006345
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a621
■1001 ▼aMichaux, Jonathan.
■24510▼aTowards Risk-Aware Planning in Uncertain Environments
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a135 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: B.
■500 ▼aAdvisor: Vasudevan, Ram.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aRobotic manipulators have the potential to enhance modern healthcare by performing diagnostic procedures in point-of-care settings, assisting surgeons in operating rooms, and accelerating the discovery of novel therapeutics in research laboratories. However, researchers must address several key challenges to ensure robots can accomplish these important tasks. First, robots should be autonomous. This means robots should be capable of sensing their surroundings, gathering and learning from new information, and making their own decisions about how to accomplish specific tasks. Second, robots should be safe and only perform actions that are guaranteed to not damage objects in the environment, nearby humans, or even the robot itself. Third, robots should be robust to uncertainty. For example, a research robot should be able to clear instruments from a workbench without knowing the exact mass or friction coefficient of each instrument. Finally, robots should operate in real-time. This ensures that robots can quickly adapt their behavior to task or environment changes.The goal of this thesis is to address the challenges discussed above by developing a novel motion planning framework that integrates perception, reachability analysis, and control algorithms to generate safe trajectories in a receding horizon fashion. The proposed framework is the result of the following major contributions: (1) the development of a core trajectory planning framework based on reachability analysis; (2) the development a novel sphere-based safety representation that facilitates the integration of perception models into the planning framework; and, (3) the application of the trajectory planner as a novel differentiable neural network layer. My approach is distinctive in three ways. First, it ensures that both safety and dynamics constraints are robust to uncertainty and satisfied in continuous-time rather than discrete-time. Second, it studies how to integrate artificial intelligence, trajectory optimization, and control into a coherent framework for robot learning with theoretical guarantees. Third, by leveraging accelerated computer hardware, the proposed framework is computationally tractable and capable of being implemented in real-time on real robotic systems. In this thesis, we demonstrate this framework's effectiveness by solving a variety of challenging motion planning and manipulation tasks in simulation and on real hardware.The first major contribution of this thesis is the development of an algorithm called Autonomous Robust Manipulation via Optimization with Uncertainty-aware Reachability, or ARMOUR. As we show throughout this thesis, ARMOUR is the core algorithm of the proposed planning framework. The key insight behind ARMOUR is to combine classical robotics algorithms with interval arithmetic to compute reachable sets that overapproximate the behavior of the robot in continuous-time. This allows one to use the forward kinematics to overapproximate the swept volume of the robot and inverse dynamics to overapproximate its dynamics. ARMOUR integrates a technique called reachability analysis with a novel nonlinear robust controller to generate safe trajectories that a robot follows in successive steps to reach a goal state. At every planning iteration, a continuum of parameterized trajectories is used to construct reachable sets that overapproximate all of the robot's positions, velocities, and torques that can be reached from a given set of initial conditions over a specified interval of time. To account for uncertainty in the robot dynamics, the position reachable set is buffered by the robust controller's worst-case tracking error. Using the reachable sets as constraints, ARMOUR solves a nonlinear optimization problem to find a trajectory that brings the robot close to a desired goal, is guaranteed to avoid collisions with obstacles, and satisfies the position, velocity, and torque limits of the robot. ARMOUR is demonstrated to outperform other state-of-the-art motion planning algorithms such as ARMTD and CHOMP. The second major contribution of this thesis addresses two key limitations of ARMOUR by developing of an algorithm called Safe Planning for Articulated Robots using Reachability-based Obstacle avoidance With Spheres, or SPARROWS. First, the obstacle-avoidance reachable sets constructed by ARMOUR are overly conservative. This can make it difficult for ARMOUR to generate motion plans in cluttered scenes. Second, ARMOUR assumes all obstacles in the scene are known and can be modelled as convex polytopes. To overcome these limitations, SPARROWS develops a novel sphere-based representation that is shown to be far less conservative than ARMOUR. This results in more flexible, less conservative planning in cluttered environments, significantly outperforming additional state-of-the-art motion planners such as TrajOpt, MPOT, and cuRoboFinally, this thesis introduces SPLANNING, a perception-based, risk-aware trajectory planning framework. Unlike ARMOUR and SPARROWS, SPLANNING represents complex scenes as collections of normalized 3D Gaussian splats. This representation provides a smooth and fully differentiable description of the workspace, avoiding the limitations of point clouds and occupancy grids. SPLANNING combines the 3D Gaussians with the spherical reachable sets from SPARROWS to compute an upper bound on the probability of collision between the robot and its environment. This bound is then used as a constraint in a gradient-based optimization framework, enabling real-time planning in visually complex scenarios. Together, ARMOUR, SPARROWS, and SPLANNING form the first reachability-based framework that unifies perception, planning, and control into a general approach for safe and efficient manipulation under uncertainty.
■590 ▼aSchool code: 0127.
■650 4▼aMechanical engineering
■650 4▼aRobotics
■650 4▼aComputer engineering
■653 ▼aRobots
■653 ▼aMotion planning
■653 ▼aConvex polytopes
■653 ▼aPolynomial zonotopes
■653 ▼aManipulation
■690 ▼a0771
■690 ▼a0800
■690 ▼a0548
■690 ▼a0464
■71020▼aUniversity of Michigan▼bRobotics.
■7730 ▼tDissertations Abstracts International▼g87-03B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359843▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


